Abstract: A method of generating responses responsive to input prompts including receiving a multimodal input prompt, the multi-modal input prompt comprising a first input prompt mode and a second input prompt mode, generating first and second derived prompts responsive to the first and second input prompt modes, transmitting the first and second derived prompts to first and second h-LLMs that are trained using data having a mode that is the same of the respective input prompt modes, receiving one or more first and second results from the first and second h-LLMs, generating at least one combined result from the one or more first and second results, and transmitting the combined result to the user.

Abstract: A method of generating derived prompts including receiving a user prompt at a user interface and generating a plurality of derived prompts from the user prompt at an input broker.

Abstract: A method for training large language models (LLMs) and using them for inference including training a base LLM, creating connected models while training the base LLM by coupling multiple trained base a parallel, series, or hybrid architecture, creating specialized connected language models for different specialized processing tasks by supplementally training two or more connected models different respective training sets, supplementally training a subset of connected language models by selectively routing one or more training data inputs to one or more specialized connected language models of the plurality of specialized connected language models responsive to at least one of accuracy optimization and task specialization, and routing prompts or derived prompts to the subset.

Abstract: A method of answering queries using one or more families of large language models (h-LLMs) including receiving a user prompt via an application programming interface (API) at an input broker from a user interface, generating a plurality of derived prompts from the user prompt at the input broker, transmitting the plurality of derived prompts to one or more h-LLMs via a cloud service API, each h-LLM of the one or more h-LLMs being in a cloud container environment, receiving a plurality of h-LLM results at an output broker, processing the plurality of LLM results at the output broker to generate a result, and transmitting the result to the user interface via the API.

Abstract: A method for improving performance of functions-as-a-service including receiving at a function delivery network a first function call from a first user and a second function call from a second user calling the first function within a first time window. The method further includes sending a first function call request to a cloud service responsive to the first and second function calls, performing at the cloud service a first function responsive to the first function call request producing a first function result, receiving at the function delivery network the first function result, and transmitting from the function delivery network the first function result to each of the first user and the second user.

Abstract: A method of generating and searching for labels in a document file including receiving a document file, extracting extracted text from text elements in the document file, indexing the extracted text to produce a list of extracted labels, receiving a request to open the document file from a user browser, sending the document file to be displayed on the user browser, receiving an area selection on an area of the document file displayed on the user browser from the user browser, identifying a selected label located within the area selection, locating a matching label from the list of extracted labels, and sending each of the matching labels to the user browser for display on the user browser.

Abstract: A method of generating and searching for labels in a document file including receiving a document file, extracting extracted text from text elements in the document file, indexing the extracted text to produce a list of extracted labels, receiving a request to open the document file from a user browser, sending the document file to be displayed on the user browser, receiving an area selection on an area of the document file displayed on the user browser from the user browser, identifying a selected label located within the area selection, locating a matching label from the list of extracted labels, and sending each of the matching labels to the user browser for display on the user browser.

Abstract: A method for removing unauthorized information associations from an LLM including identifying real UD instances training data, identifying UD associations for the UD instances, generating synthetic UD instance-UD association pairs to reduce or remove influence of a real UD instance-UD association pair on an output of the LLM, an adversarial dataset including synthetic UD instance-UD association pairs and the real UD instance-UD association pairs, and generating a fine-tuned LLM by iteratively fine-tuning the LLM using the adversarial dataset.

Abstract: A method for improving responses to large language model (LLM) prompts including receiving at an input broker a request input from a user including an LLM prompt, deriving a search query from the LLM prompt, searching a plurality of documents using the search query to identify a first subset of documents and a second subset of documents, generating a first answer using a first context-specific LLM, where the first context-specific LLM uses the first subset of documents as the context and the LLM prompt as the prompt, generating a second answer using a second context-specific LLM, where the second context-specific LLM uses the second subset of documents as the context and the LLM prompt as the prompt, providing each of the first answer and the second answer to an output broker, determining a primary result at the output broker, and transmitting the primary result to the user.

Abstract: A method for training large language models (LLMs) and using them for inference including training a base LLM, creating connected models while training the base LLM by coupling multiple trained base a parallel, series, or hybrid architecture, creating specialized connected language models for different specialized processing tasks by supplementally training two or more connected models different respective training sets, supplementally training a subset of connected language models by selectively routing one or more training data inputs to one or more specialized connected language models of the plurality of specialized connected language models responsive to at least one of accuracy optimization and task specialization, and routing prompts or derived prompts to the subset.

Abstract: A system for enhancing large language models (LLMs) including an input component to receive an input text and split the input text into tokens, a batch processing component including one or more LLMs fine-tuned based on an aggregation of input data and that generate outputs responsive to receiving a token, a ranking component to score the LLM outputs and rank the tokens responsive to their respective scores, a clustering component to select a subset of highest-ranked tokens and consolidate the subset into refined context token batches, a control component to provide the refined context token batches for further training of the LLMs, and a query component to receive input queries, generate derived queries responsive to the input queries, transmit the derived queries to the batch processing component, receive responses from the batch processing component responsive to the derived queries, score and/or rank the responses, and transmit one or more responses.

Abstract: A method of facilitating code development comprising receiving a code file from and creating or updating a meta-file comprising code content and metadata for the code file. The code file may be maintained in a code repository and the metadata in a metadata storage. The information contained within the metadata may convey design decisions, architectural guidelines, and other parameters governing code development. The code file and the metadata content may be displayed on a user device in an IDE or code editor.

Abstract: A method of generating a certificate smart contract including receiving a certificate smart contract transaction to create a certificate smart contract from an owner of a blockchain wallet, generating a certificate smart contract responsive to the certificate smart contract transaction at a certificate smart contract address on a blockchain network, receiving a new certificate transaction to create a new certificate to be included by the certificate smart contract from a validator, and saving a certificate to the certificate smart contract responsive to the new certificate transaction. The certificate smart contract includes an identity smart contract.

Abstract: A method for detecting fraud across networks of interconnected smart contracts on a blockchain network, including constructing a graph representation of nodes and node edges representing smart contracts and interactions therebetween on a blockchain network. New smart contracts are added to the graph representation which is updated to reflect updated edges. Pattern recognition algorithms are applied to the graph database to identify and flag sub-graphs suggesting potential fraud. Filtering smart contracts gather forensic data about each smart contract comprised by the flagged sub-graph. A fraud risk score is determined by analyzing the forensic data. If the fraud risk score exceeds a threshold value, a fraud response is triggered.

Abstract: A cross-blockchain value transfer approval method including the steps of receiving a cross-blockchain value transfer request command that includes a first address on a first blockchain network, a first cryptocurrency type indication associated with the first blockchain network, a second address on a second blockchain network, a second cryptocurrency type indication associated with the second blockchain network, and a transfer quantity. The method further includes sending a transaction approval request to a second user, receiving a cross-chain respond command API request from the second user, and upon receiving the cross-chain response command approving the transfer, sending a command to initiate a value transfer to an exchange server.

Abstract: A method of answering queries from user prompts comprising generating derived prompts from a prompt, generating prompt embeddings from the derived prompts, transmitting the prompt embeddings to a vector database, receiving knowledge documents from the vector database, generating context-aware prompts responsive to the user prompt, the derived prompts, and the knowledge documents, transmitting the context-aware prompts to h-LLMs, receiving h-LLM results from the h-LLMs, processing the h-LLM results to produce processed h-LLM results, each processed h-LLM result having a score, identifying preferred results of the processed h-LLM results having the greatest scores, and transmitting the one or more preferred results to the user.

Abstract: A method for detecting fraud across networks of interconnected smart contracts on a blockchain network, including constructing a graph representation of nodes and node edges representing smart contracts and interactions therebetween on a blockchain network. New smart contracts are added to the graph representation which is updated to reflect updated edges. Pattern recognition algorithms are applied to the graph database to identify and flag sub-graphs suggesting potential fraud. Filtering smart contracts gather forensic data about each smart contract comprised by the flagged sub-graph. A fraud risk score is determined by analyzing the forensic data. If the fraud risk score exceeds a threshold value, a fraud response is triggered.

Abstract: A cross-blockchain value transfer approval method including the steps of receiving a cross-blockchain value transfer request command that includes a first address on a first blockchain network, a first cryptocurrency type indication associated with the first blockchain network, a second address on a second blockchain network, a second cryptocurrency type indication associated with the second blockchain network, and a transfer quantity. The method further includes sending a transaction approval request to a second user, receiving a cross-chain respond command API request from the second user, and upon receiving the cross-chain response command approving the transfer, sending a command to initiate a value transfer to an exchange server.

Abstract: A blockchain transaction filtering method including receiving a transaction request at the server, executing a first smart contract function comprised by a first smart contract stored on the server responsive to the transaction request, executing a first filter smart contract function comprised by a first filter smart contract stored on the server responsive to the transaction request, the first filter smart contract function checking the transaction request for inconsistency with a first filtering criterion, defining a first identified transaction request and implementing a first response responsive to identifying the first identified transaction request.

Abstract: A method for implementing zero-knowledge private key management for decentralized applications including receiving a session request, establishing a session responsive to the session request, transmitting a response to the session request to the decentralized application, receiving a session approval from a client application, updating the session with the information included in the session approval, and transmitting the public key and the blockchain network selection to the decentralized application.